Extended DownloaderFactory to support optional username & password parameters.
Extended HttpDownloader to check for HTTP 401 Authorization Required status code
and send a simple HTTP Basic Authentication header with all requests.
Extended ManageReposActivity to support repositories that use HTTP Basic
Authentication, added a dialog to prompt for username and password.
Extended RepoDetailsActivity to be able to display and modify the authentication
credentials.
Replace search dialog with a search widget
SearchView is the recommended way to implement search UI. See https://developer.android.com/guide/topics/search/search-dialog.html#UsingSearchWidget
The UX is still far from ideal but looks much better now.
Before and after (Gingerbred):
Before and after (Lollipop):
See merge request !168
Not sure why it was added initially but now it appears to be unneeded:
the support library does everything right and the lists are themed
properly without any hacks.
Our `IconDownloader` extended `BaseImageDownloader` from UIL. There was an
explicit check in the F-Droid `IconDownloader` which looks for
HTTP/HTTPS/Bluetooth schemes. If it wasn't one of these, it fell back
to the base class. This was what was happening for local cached image
files. As such, when the `getInputStream(...)` method was refactored
to only use F-Droids `DownloadFactory` and not delegate to the base class,
it failed on local "file://" URLs.
This change introduces a `LocalFileDownloader` and makes the `DownloaderFactory`
aware of it.
The `BaseImageDownloader` class only provides support for the following schemes:
* HTTP
* HTTPS
* File
* Android content providers
* Android assets
* Android drawables
F-Droid now supports HTTP, HTTPS, and File URLs. There is not currently any
need for content proiders, assets or drawables to get icons for apps in F-Droid.
If there is a need in the future (e.g. an issue currently discusses loading
icons from installed apps if possible) then that specific `Downloader` can get
introduced to solve the problem.
These are written manually and mostly don't contain HTML. Some html is
fine if you want to use links or markup, but <p> elements are just
pointless and very seldom used. Be consistent in not using them.
Added tests for multiple repositories providing same apks
Right now, multi repo support works, but is kinda funky. While fixing #324, I accidentally broke this support without realising it. So in the interests of making my approach to #324 more test driven, I've written some tests for multi repo support.
Initially I wrote tests for the actual correct, desirable behaviour. Then when it became apparant that we dont' do this, I commented those tests out (but left them there for hopefully future multi-repo work) and then added tests for the current behaviour to make sure we don't introduce regressions.
Android unit testing framework is nice for testing content providers. It is nice for testing file handling. However I really struggled to get it working with both. Had to do some interesting things with instrumentation and contexts in order to get it to work.
I'm sure Android has nice `Service` testing capabilities too. But given the trouble with instrumentation/contexts/files/providers/etc, it was easier for me to refactor the parts of `UpdateService` that I needed to test into a separate, testable class.
See merge request !163
Came across this whibluetooth swap.
I think it is reasonable to guard against null activities here, because
we are likely not releasing observers correctly, and thus the observer
may receieve a notification when the activity is not attached.
The base `Downloader` class now wraps the `InputStream` returned by
any child classes, in order to notify the child class when that stream
is closed. This prevents each child class having to figure out a way
to be notified of this.
Also removed some dead code which was unused.
(One of) the problems with F-Droid's multiple support is that there is
a primary key on the fdroid_apk table which is a composite of:
* id
* vercode
Which means that two repos providing the same version means one will
update the other, rather than ending up with two different versions.
Instead, there should be some other way to differentiate apks from
different sources. Firstly, it should take into account the signing
cert. Secondly, it may taken into account the hash, because two people
could sign different apks with the same cert and then we are back at
square one.
Make category filter look Material
Get rid of Gingerbred legacy.
Before and after (note the glitch below the spinner):
See merge request !159
We were fetching information on all installed packages and doing a linear
search. Which is silly and inefficient since we can directly fetch information
on a single installed package by id.
If it doesn't result in any version being marked as suggested, e.g. because it
is higher than 0 but lower than any available version, we end up with a NULL
suggestedVercode.
Run the Latest() algorithm after the Upstream() one, and have it pick up those
cases too by adding "OR suggestedVercode IS NULL". This way, we treat invalid
upstreamVercode values as if they were NULL.
Also some cleanup of comments.
Fixes#371.
